<?xml version="1.0" standalone="yes"?>
<Paper uid="C92-3156">
  <Title>Agent AGT Beneficiary BENF Experiencer EXPR Instrument INST Object OBJ Recipient RECP Direction DIR Location_at LAT Location_from LFRM Location_to LTO Location_through LTRU Orientation ORNT Frequency FREQ Time_at TAT Time_from TFRM Time_to 'ITO Time_through qTRU Cause CAUS Contradiction CNTR Effect EFF Purpose PURP Accompaniment ACMP Content CONT Manner MANR Material MATR Measure MEAS</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
CONCLUSION
</SectionTitle>
    <Paragraph position="0"> We have presented the DIPETI' parser and the Case Analyzer--the main elements of the linguistic part of the TANKA system. TANKA will process unedited technical text and acquire knowledge about its domain. We want to analyze complete documents with as little user assistance as possible. This means that we must consider incomplete and problematic inputs, although their rate of occurrence should be low in a welledited text. We have ensured robust low-level processing of text in order to facilitate almost automatic recognition of its structure. At the other end of the spectrum, we plan to handle free segments of text. In contrast with other approaches to language understanding, we do not assume a complete semantic model apriori. This imposes certain limitations on what can be processed automatically; we will minimize user interaction. We hope that we have made clear our interest in practical NLP, which we regard as important given the increasing interest in using NLP techniques to assist in acquiring knowledge from text. We believe such techniques will be used more and more commonly for knowledge acquisition tasks and may establish a new trend in the design of tools for knowledge engineers.</Paragraph>
  </Section>
class="xml-element"></Paper>